USE OF LINEAR REGRESSION METHOD FOR CRYPTOCURRENCY PRICE PREDICTION

Abstract

Predicting cryptocurrency prices is one of the most researched topics and is of interest both scientifically and commercially. In the past, traders used various complex formulas and charts for forecasting, analyzed them, and based on their findings, built trading strategies. Now, with the advent of artificial intelligence, various algorithms have been employed to predict stock market movements, including various cryptocurrencies.

Machine learning algorithms are used both for short-term stock price prediction and for long-term market understanding. At present, there are quite a few studies in the public domain on the use of artificial intelligence for predicting markets with low volatility, such as indices or international company stocks. However, there are still few studies describing the prediction of high-volatility markets, such as the cryptocurrency market. These markets are known for their unpredictability and complexity, making them interesting for researchers and traders.

This article explores the application of the linear regression method for predicting prices in a high-volatility stock market, such as the cryptocurrency market. For this study, the cryptocurrency Ethereum was chosen. The author analyzes the impact of various input data, such as price history, trading volumes, and other technical indicators, on the model's performance. Additionally, the impact of different model parameters on the quality of results and training time is compared.

Moreover, a trading strategy based on the model's predictions is proposed, and the results of its performance in a cryptocurrency market simulation are described. The analysis of the results shows the advantages and disadvantages of using linear regression for predicting high-volatility markets. In conclusion, the author determines the feasibility of using such a model for real trading and suggests ways to improve prediction accuracy.